To investigate whether Prdm1 acts as an activator or repressor of transcription during slow-twitch muscle development in zebrafish, constructs in which the Prdm1 DNA-binding domain is fused to either the Engrailed (Eng) repressor or the VP16 activator domain (Kessler, 1997) were tested for their ability to substitute for the wild-type protein. Sequences encoding the fusion constructs were cloned downstream from a heat-shock-inducible promoter that simultaneously drives expression of the fluorescent protein tdTomato. Transient expression of the Eng–Prdm1 protein induced by heat shock was sufficient to rescue the expression of Prox1 in ubo^tp39 mutant embryos (Fig 2C). By contrast, transient expression of the Vp16–Prdm1 fusion protein was unable to rescue Prox1 expression or suppress myoblast fusion (Fig 2D,E), but was sufficient to activate fast-myosin expression. Taken together, these data indicate that Prdm1 acts as a repressor rather than an activator to promote slow-muscle differentiation, and show that activation of prox1 and smyhc1 expression is an indirect effect of Prdm1 activity. As the PR domain is absent from the Eng–Prdm1 protein fusion, it also follows that this domain is not essential for Prdm1 function, at least in this context. Consistent with this, previous studies have shown the PR domain to be dispensable for Prdm1 function in the context of β-interferon repression (Gyory et al, 2004).

In mice, the transcription factor Sox6 acts as a repressor of fetal slow-twitch fibre differentiation (Hagiwara et al, 2005, 2007). We analysed expression of the zebrafish sox6 gene and found that it was expressed in the fast-muscle progenitor domain of the somites but excluded from adaxial cells (Fig 3A). In ubo^tp39 mutant embryos, by contrast, sox6 is ectopically expressed in adaxial cells, indicating that Prdm1 represses sox6 expression in slow-muscle progenitors. Forcing ectopic expression of sox6 in the adaxial cells of wild-type embryos caused an inhibition of Prox1 expression (Fig 3B). Conversely, morpholino-mediated knockdown of sox6 in ubo^tp39 embryos partly rescued Prox1 expression and restored expression of smyhc1 to normal levels in adaxial cells (Fig 3C). Although neither Prox1 nor smyhc1 was expressed ectopically in fast fibres in response to sox6 knockdown, robust expression of stnnC was induced in the fast muscle of both wild-type and smoothened (smo) mutant (that lack all Hedgehog signalling activity) embryos injected with the sox6 morpholino (Fig 3E,F). This disparity might reflect a differential sensitivity of these slow-twitch-specific genes to sox6 activity, revealed by incomplete knockdown by the morpholino. Taken together, these data indicate that Sox6 acts as a repressor of slow-twitch-specific gene expression and suggest that Prdm1 activates such expression by repressing transcription of sox6.

The coordinated repression of multiple genes encoding fast-twitch isoforms of sarcomeric proteins could be accomplished by Prdm1-mediated repression of a fast-specific global transcriptional activator; alternatively, Prdm1 might itself act directly to repress transcription of these genes. To distinguish between these two scenarios, we used chromatin immunoprecipitation (ChIP) to test for binding of the protein to upstream regulatory regions of the putative targets. A polyclonal antibody was raised against 186 amino acids from the Prdm1 PR domain; as expected, this labelled adaxial cell nuclei during early- and mid-myogenesis (Fig 4A), and bound to Prdm1 specifically in immunoprecipitation assays (Fig 4B). DNA from chromatin immunoprecipitated with this antibody was amplified by using primer pairs specific for sequences proximal to the transcription initiation sites of the mylz2, fMyHCx, tnnt3a and tnni2 genes. These sequences were all found to be enriched in the Prdm1-precipitated chromatin; by contrast, none of the slow-muscle-specific genes smyhc1, stnnC or Prox1 was enriched (Fig 4C). These data indicate that Prdm1 selectively binds to putative regulatory regions of fast-fibre-specific genes in vivo, suggesting that Prdm1 acts as a direct repressor of their transcription.

Expression of a GFP reporter gene containing 2.3 kb of the mylz2 promoter sequence is specifically repressed in adaxial cells by Prdm1 activity (Fig 4D). Although the consensus binding site for Prdm1 has not been determined in zebrafish, we identified five putative Prdm1-binding sites in this fragment, containing the GAAAG core of the sequence (A/C)AG(T/C)GAAAG(T/C)(T/G) that has been defined as mediating Prdm1-dependent gene regulation in mammals (Kuo & Calame, 2004). The introduction of point mutations in each of these five potential Prdm1-binding sites in this construct led to ectopic adaxial GFP expression in wild-type embryos, similar to that seen with the wild-type construct in ubo morphants (Fig 4D). This finding is consistent with Prdm1 acting directly to repress the mylz2 gene in adaxial cells at the 12-somite stage.

To confirm and extend the findings of our candidate gene analysis, we used a recently constructed zebrafish promoter array, consisting of 60-mer probes for more than 11,000 genes within the zebrafish genome (Wardle et al, 2006), to probe the DNA isolated by ChIP of myoblast extracts (supplementary information online, accession code GSE10883 at GEO). By setting the gene array threshold for enrichment to the level of mylz2, which was at the level of significant P-value 0.0078 and P[Xbar] 0.0075, we identified 381 putative target genes (supplementary information online). Gene ontology analysis showed various genes enriched in the ChIP-on-chip sample: 11% were documented transcription factors, 24% were new genes or genes without known function, 15% were genes encoding proteins with enzyme activity, such as kinases and phosphatases, whereas others had gene ontology terms linking them to the immune and haematopoietic systems, cell-cycle regulation or apoptosis. Significantly, we found several genes encoding fast-fibre-specific isoforms of sarcomeric proteins, including those encoding fast MyHC and troponins described above, whereas no genes encoding slow-specific sarcomeric proteins or Prox1 were identified (supplementary information online). Surprisingly, sox6 was not among the transcription factor-encoding genes identified in this analysis. However, we note that representation of regulatory regions on the gene array is restricted to sequences 9 kb upstream from the 5′ end of the complementary DNAs used in its design (Wardle et al, 2006). We have identified additional sox6 sequences 30 kb upstream from the transcription start site used in the array (J.v.H., S.E. & P.W.I., unpublished data); whether sox6 is a direct target of Prdm1 remains to be determined.